AI Art Today

As an American living in the digital age, I’m constantly exposed to artificial intelligence. It has infiltrated everything, including social media platforms, marketing, and search engines. And this is no surprise–AI does offer its share of financial incentives. That’s why it’s so interesting that there’s one use of AI that doesn’t explicitly have any financial incentive: AI “art”. While I know that there are people that sell it, there are communities that generate and share AI art just as people show off art they create themselves. As an artist myself, I’m a little confused by this. Let’s take a look at how we got here, the controversy of AI art, and some insight into how society should move forward.

How did we get here?

I’m not going to go too in depth with the history of generative AI art because it gets a little boring even for me as a computer science major. For our purposes, let’s go back to the distant year of 2021. This was the year the world saw the explosion of text-to-image models like OpenAI’s DALL-E, in which users could input a description of an image and the model would output a few instances of pictures that fit the description. They were very distorted and strange, but funny.

Image via Arts Management and Technology Laboratory

With the rise in popularity of AI images, more models and improvements came about like Midjourney, Stable Diffusion, and DALL-E 2. The differences, even in such a short period, are staggering. AI generated images have changed a lot in just a few years and the progress isn’t stopping. As these technologies develop and become more well-known, the discourse surrounding them has grown as well.

Image via Medium

Why do people like AI art?

Accessibility is something that gets brought up a lot in pro-AI art spaces. There’s also the argument for affordability–having a computer make something for you is much cheaper than producing it yourself. And I get it: I have invested a lot of time and money into obtaining my art supplies and learning how to use them. Sometimes it’s challenging and sometimes it’s frustrating, but that’s part of the process.

I feel like there’s this idea that people who don’t take the time to develop the skills to paint, draw, or illustrate are entitled to create art in a similar vein as those who have developed such skills. This devalues the effort of those who have put in that work. And to the point of financial strain,  art could also be more accessible if there was more appreciation and compensation for those in artistic fields. 

Why are people against it?

For one, like all AI generation, the creation of AI images takes a lot of power. Generating just one image can use as much energy as charging a phone. The process emits carbon and requires a lot of freshwater to cool down the hardware. This alone is terrible and a very valid reason to hate AI art, but this also isn’t specific to just art.

Image via The Washington Post

Image via The Washington Post

Another issue that’s inherent to AI but is particularly evident in its image generation is the presence of bias. When AI models are trained, they ‘learn’ from whatever information they are fed. When this information is biased, the bias becomes a part of the model. This has led to some pretty atrocious outcomes.

Depicted above are images generated by Stable Diffusion when prompted to depict an “attractive people” and “toys in Iraq,” respectively. Notice how in the first set of images, everyone is white. In the second, every toy is a soldier brandishing a gun. In another example, AI overwhelmingly depicts people with high-paying jobs as white or light-skinned and people with low-paying jobs as people of color. These biases are present ingrained into AI are a problem and using it to generate art only makes it worse.

But what makes AI ‘art’ particularly controversial is how it affects artists. The way AI art is generated is essentially by consuming images, learning patterns, and then using them to ‘produce’ something. This means that the work it’s trained on was made by someone else and its product is an amalgamation of stolen work. AI companies get their training data by scraping the internet–taking images made by artists and feeding it to the machine to use for parts. Artists have little control over or protections for their work.

This means that companies can produce replicas or ‘new’ content (albeit, not often anything good) without providing any compensation to the original creator or without contracting an artist to create something original. It cuts into artists’ livelihood and greatly devalues their work. Professional designer Matt Corrall puts it nicely: “‘AI art’ is something very different. Boundary-pushing art never threatened the careers of other artists before, or threatened to monopolize the means of creation - it simply challenged our ideas about ourselves and our world.” Art can mean a lot of vastly different things. From my perspective, it takes skill, time, and thoughtfulness; AI ‘art’ doesn’t require any of these. Companies involved with this theft are unjustly capitalizing off other’s labor.

It just begs the question: why art? Art is a profoundly human endeavor and, believe it or not, many people (like myself) actually enjoy making it! So why mechanize art instead of the menial, thoughtless tasks that no one actually wants to do? 

The irony is that while these big tech companies try to push humans out of art, they’re inadvertently limiting their data pool. As the internet becomes bogged down with AI content, the models are scraping it and feeding it back to themselves for training. 

How should we move forward?

For legislators:

I know that it’s hard for legislation to keep up with the rapid advancement of AI, but something has to be done. Measures need to be taken to protect artists from having their property stolen and used in AI training. Most smaller artists don’t have the resources to enter a legal battle with these large AI companies, and there needs to be something to keep them in line.

For businesses:

While you may think you’re saving money by not hiring a designer, you might actually be hindering your profits. In one study, researchers found that AI generated advertisements induced feelings of discomfort and doubt about the brand’s authenticity in about half of consumers. Personally, it comes across like they couldn’t afford to hire a designer. 

For artists:

Protect yourselves! While not 100% perfect solutions, there are tools popping up that can help limit the AI’s ability to be trained on your art. Glaze is one method to prevent theft. It changes the art almost imperceptibly to humans but drastically to AI. It makes it so that certain styles are difficult to be recognized and replicated by AI. It’s a defensive method used to prevent mimicry. 

Similarly, Nightshade is a tool that is designed to “poison” art used to train without the owner’s consent. It slightly modifies the pixels of the images so that AI sees it as something totally different than what it is. Nightshade can withstand changes like cropping, noise, and compression as well.

For everyone:

Don’t engage with AI art. Don’t share it, interact with it, or boost it. It’s becoming harder to spot, but there are still giveaways: strange texture changes, weird blurs, blank stares, and slightly off anatomy are all things to look out for.

It’s obvious that art is a valuable facet of humanity, and AI completely disregards that. It’s becoming more challenging for artists to support themselves now that AI is devaluing their efforts. If you are able, find ways to support artists, especially local ones. Commissioning someone if you find a piece you really like can help out a lot.

Finally, engage with and participate in the arts! See for yourself how rewarding and frustrating and emotional and empowering it is. Your experiences and perspectives are in every dot and line you draw. Take pride in your ability (even if you’re not very good, that’s okay!). What you create as a human is inherently more valuable than the art produced by a machine.